Guiding Generative Protein Language Models with Reinforcement Learning

TL;DR

This paper introduces a reinforcement learning framework to guide protein language models in designing proteins with specific desired properties, significantly improving their ability to generate high-fitness variants efficiently.

Contribution

It presents a novel method combining reinforcement learning with protein language models to steer their outputs toward user-defined objectives, enabling rapid and targeted protein design.

Findings

Achieved a 26-fold increase in EGFR binder affinity in two iterations.

Successfully guided pLMs toward various protein properties such as topology and binding affinity.

Demonstrated efficient design with few iterations through evolutionary trajectories.

Abstract

Protein language models (pLMs) have demonstrated success at generating functional proteins across vast sequence spaces but lack the ability to design high-fitness variants on demand. Here, we iteratively guide pLMs toward user-defined objectives by applying reinforcement learning (RL). We demonstrate that RL can steer pLMs toward various protein properties, such as topologies or binding affinities, in a few iterations through long evolutionary trajectories. We apply our framework to the design of epidermal growth factor receptor (EGFR) binders, achieving a 26-fold increase in binding affinity in two iterations.